- Now let's say you took a statistics class and your lecturer
taught you hypothesis testing and then told you that it is useless. The good
news is you can still use the tests to construct confidence intervals.
Larry Wasserman, Intermediate Statistics
- Each of you will get their own set of data and calculate
confidence intervals. Not everyone of you will get lucky -- some of you
will miss the true value of theta -- that's life. That's how statistics
works. But 95% of you will trap the true value. That's the best we can
hope for.
Larry Wasserman, Intermediate Statistics
- Power of a test is about the probability of winning the Nobel prize.
Larry Wasserman, Intermediate Statistics
- We are gonna cover permutation test, because there
will be a time when you will say to yourself "Man, am I glad he
taught us that."
Larry Wasserman, Intermediate Statistics
- There is a lot of theory developed for it in 70s.
But... that's not that interesting.
Larry Wasserman, Intermediate Statistics
On robustness.
- There is a lot of probability out there. Far away.
Larry Wasserman, Intermediate Statistics
On Cauchy distribution.
- Student: Why is it so?
Larry: Because of this theorem.
Student: ?
Larry: It has a proof.
Student: It's so clear now.
Larry: Just go 5 times through the proof.
Larry Wasserman, Intermediate Statistics
-
Just look at the numbers and do the intuition in retrospect.
S. Rudich, "Great Theoretical Ideas in CS" (on probability)
-
It happened with me three times when I was told "There's no way you can solve it." Once it was a 25-year old problem in graph theory. There was a bunch of Hungarians thinking over it and they told me: "Don't waste your time over dinner." By the end of the dinner I solved it.
S. Rudich, "Great Theoretical Ideas in CS"
-
It's enough to do something like this three times in one's lifetime and they will be saying:"He...whoa."
S. Rudich, "Great Theoretical Ideas in CS"
-
How to solve the differential equation? Gee, there are professors who do it for living.
S. Rudich, "Great Theoretical Ideas in CS"
-
The manager asked him "Have you sexually molested our client?", and he replied "Yes, I did." Now, Bill Clinton could...
S. Rudich, "Great Theoretical Ideas in CS"
-
People played with Pascal's Triangle and produced a lot of results (that's before masturbation was invented).
S. Rudich, "Great Theoretical Ideas in CS"
- Why do we call it "Pascal's Triangle"? Al-Karaji who lived in Baghdad (953-1029) knew it, Chu Shin-Chieh knew it in 1303, and it appeared in The Precious Mirror of the Four Elements. It was known in Europe by 1529, and to Blaise Pascal in 1654. Can you pronounce those names? That's why it's called "Pascal's Triangle".
S. Rudich, "Great Theoretical Ideas in CS"
- Some other majors have their mystery numbers, like \pi and e. In Computer Science the mystery number is \phi, the Golden Ratio.
S. Rudich, "Great Theoretical Ideas in CS"
- Fibonacci has become interested in rabbits and what they do.
S. Rudich, "Great Theoretical Ideas in CS"
- When you are multiplying two 60-digit numbers Karatsuba starts to look good.
S. Rudich, "Great Theoretical Ideas in CS"
- I thought Karatsuba was dead, but then I've got some email.
S. Rudich, "Great Theoretical Ideas in CS"
- This lecture has explicit mathematical content that can be shocking to some students.
S. Rudich, "Great Theoretical Ideas in CS"
- To some guys anything female is basically OK.
S. Rudich, "Great Theoretical Ideas in CS"
- And the curse of dimension doesn't bother us anymore.
S. Smale
- You shouldn't be too nervous about neural networks.
Neural Networks course.
- You can find it in some digital books.
(Translation: in books on Digital Control.)
Discrete-Time Control Systems course.
- You should be careful about expectation and independence.
What would you think? It was about probability. DSP course.
- This method is OK if the function is not too complex.
Tricky one again. Imagine if it was about a value of imaginery
part of the function. It wasn't. DSP course.